An integrated circuit overlay error measurement method based on information fusion

Abstract

The application provides an integrated circuit overlay error measurement method based on information fusion, which is characterized in that, for a given diffraction or scattering type overlay error measurement mark, model-based overlay error measurement (MBO), experience-based overlay error measurement (EBO), model / model hybrid measurement (MBO+MBO) and model / experience hybrid measurement (EBO+MBO) are used to extract overlay error values. Based on the calculation of electromagnetic field simulation means, the measurement distribution and its uncertainty of different methods are estimated, the prior knowledge is provided to eliminate abnormal measurement results, and the fusion decision basis is provided. Further, the different measurement extraction calculation results obtained by the actual measurement signal are fused by using a hierarchical information fusion method to obtain the final overlay error result. The method provided by the application can realize high-credibility extraction and calculation of overlay error, and is suitable for overlay error measurement under non-ideal measurement conditions.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of integrated circuit front-end measurement, in particular to an integrated circuit overlay error measurement method based on information fusion. BACKGROUND

[0002] In integrated circuit manufacturing, lithography is one of the most complex and critical process steps, and overlay error is an important dimensional parameter in lithography process. Overlay error refers to the deviation of the current layer pattern relative to the reference layer pattern along the x and y directions. In actual manufacturing process, error cannot be avoided, in order to ensure that the circuits designed in the upper and lower layers can be reliably connected, the overlay error value of the current layer and the reference layer should be less than 1 / 3-1 / 5 of the feature line width, and the uncertainty of overlay error measurement is 10% of the allowable error. Therefore, with the continuous development of ultra-precision lithography process, the rapid and accurate measurement of overlay error has become the top priority of lithography process quality management.

[0003] In the overlay error measurement process, the measurement results of overlay error are affected by factors such as instruments, external environment and non-ideal overlay marks. The existing schemes to improve the accuracy of measuring overlay error are: 1. An feature selection method suitable for optical scattering measurement is proposed in invention patent [CN111553064A], which shortens the offline library building time in library matching method and improves the parameter extraction accuracy; 2. A method for improving lithography overlay precision is proposed in invention patent [CN115356898A], which measures the offset error of the overlay mark and compensates for the error to improve the overlay precision; 3. A method for asymmetric calibration of overlay error measurement is proposed in invention patent [CN114690584A], which changes the position of the overlay mark to obtain different measurement signals, so that the overlay error measurement is more stable, and the overlay precision is improved.

[0004] The existing methods have the following shortcomings: 1. Most of them are optimization methods for the shortcomings of a certain overlay error measurement method, and do not fuse measurement by using different extraction methods to improve overlay precision; 2. Most patents optimize the measurement method or improve the instrument for overlay error measurement, and few consider the influence of different overlay error extraction methods on the reliability of measurement results. Therefore, the overlay error measurement method based on information fusion proposed in the application uses a variety of overlay error extraction methods to fuse measurement optimization, considers the relationship between different overlay error extraction methods and the reliability of measurement results, and realizes accurate measurement of overlay error. SUMMARY

[0005] In view of the above prior art defects, the present application aims to provide an information fusion-based overlay error measurement method, aiming to solve the problem that the measurement result of the existing single overlay error measurement method may have low reliability.

[0006] To achieve the above technical purpose, the present application provides an information fusion-based integrated circuit overlay error measurement method, comprising:

[0007] (1) According to the related information of the process object to be measured, the topographic features and material properties of the overlay mark to be measured are determined, and the optical response of a group of typical DBO overlay marks is calculated by using the parameterized modeling and electromagnetic field solving method to obtain the optical signal under ideal conditions;

[0008] (2) Load the noise (ΔI or Δm) to the simulated spectral signal or Mueller matrix as the simulated measured spectral signal;

[0009] (3) Consider the model-based extraction method, the experience-based extraction method and the multi-combination extraction method of the two methods for the above signals, and calculate the noise-loaded spectral signal to obtain a plurality of different measurement values OVL m ;

[0010] (4) According to the results of batch simulation calculation, determine the reasonable value range of the overlay error distribution, and exclude the actual measurement values not in the distribution range;

[0011] (5) For different extraction methods in step (3), calculate the actual measurement signal to obtain a plurality of different measurement values OVL m , and use information fusion technology to obtain a new overlay error value OVL m ; wherein the information fusion processing has two stages: local fusion and overall fusion; wherein the local fusion technology is reflected in: for the MBO+MBO and MBO+EBO extraction method fusion measurement, respectively use the information fusion method to obtain a higher reliability overlay error value. For the results obtained by the MBO and EBO extraction methods, the average value solving method is used to obtain the overlay error value

[0012] (6) According to the secondary decision fusion technology of OVL m ' obtained by each extraction method in step (5), that is, the results of MBO, EBO, MBO+MBO, EBO+MBO and other extraction methods are fused to obtain a higher reliability overlay error value OVL * .

[0013] In summary,

[0014] An integrated circuit overlay error measurement method based on information fusion, comprising: the overlay error measurement method is based on optical scatterometry measurement technology (SCOL or DBO); the overlay error extraction method is based on optical scatterometry measurement signal; and the measured overlay mark is a set of typical DBO measurement marks. The hierarchical information fusion technology is to use evidence theory or neural network to process the measurement results obtained by different extraction methods in stages to obtain a high reliable overlay error measurement value.

[0015] The optical scatterometry measurement technology uses various spectral scatterometry instruments, and currently mainly uses angle-resolved scatterometry instruments and ellipsometers. The final optical signals can be various spectral or angular spectral information (such as light intensity distribution measured by an angle-resolved scatterometer, ellipsometric parameters measured by a general ellipsometer, or Mueller matrix ellipsometer measured Mueller matrix, etc.).

[0016] The overlay error extraction method, the empiricism based overlay (EBO) essentially finds and quantifies the empirical relationship between the change of the measured overlay mark scattering signal and the change of the overlay error, and then constructs an equation set to solve the overlay error value; and the model based overlay (MBO) regards the overlay error as a nano-sized parameter to be measured, and iteratively calculates the overlay error value by modeling and simulating the calculated electromagnetic field of the overlay mark and the model fitting of the actually measured diffraction signal. The hybrid measurement of the two is to calculate the measurement signal of the given mark by using EBO and MBO at the same time.

[0017] A set of typical DBO overlay marks, the structural features of which are: composed of two overlay mark units (each unit is composed of two layers of gratings stacked), and each mark unit introduces artificial known deviations +D and -D; wherein the empirical measurement method needs two overlay mark units, and the model-based measurement method needs one overlay mark unit, so a set of typical marks can be designed to meet the measurement needs of the two extraction methods or their hybrid methods.

[0018] The fusion measurement method needs to model and simulate the noise loading of the optical response of the given mark to obtain the simulated measurement signal, and then use different methods to obtain different measurement results, and estimate the distribution of the overlay measurement value and its uncertainty.

[0019] Overlay measurement distribution estimation, after obtaining the measured signal, used for excluding gross errors and abnormal measurements;

[0020] The hierarchical information fusion technology firstly performs local information fusion on the result of the proposed hybrid measurement method, and then performs overall secondary fusion on the different types of overlay error extraction methods.

[0021] The local fusion technology obtains a higher overlay error value with higher reliability for the MBO+MBO and MBO+EBO extraction methods by using the information fusion method.

[0022] The overall secondary fusion obtains a higher overlay error value with higher reliability by performing decision fusion on the extraction and calculation results of the MBO, EBO, MBO+MBO and EBO+MBO methods.

[0023] The information fusion-based integrated circuit overlay error measurement method provided by the application has the following specific advantages: the influence of extreme abnormality is reduced, the overlay error data obtained by different methods are estimated according to prior knowledge, the information fusion technology is used to optimize the decision of the results of multiple overlay error extraction methods, and the reliability of overlay measurement is improved. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 The overlay error measurement method of the application is a flowchart.

[0025] Figure 2 The application is a set of typical overlay mark structure schematic diagram ideal mark structure cross-sectional schematic diagram.

[0026] Figure 3 The application is a set of typical overlay mark structure schematic diagram non-ideal mark (with side inclination angle) structure cross-sectional schematic diagram.

[0027] Figure 4 The application is a set of typical overlay mark structure schematic diagram typical set of mark top view schematic diagram.

[0028] Figure 5 The application is a set of typical overlay mark structure schematic diagram non-ideal mark (with side inclination angle) structure cross-sectional schematic diagram.

[0029] Figure 6 The application is a set of typical overlay mark structure schematic diagram non-ideal mark (with side inclination angle) structure cross-sectional schematic diagram. DETAILED DESCRIPTION

[0030] In order to make the technical scheme of the application clearer, the following specific examples will be further described. The information fusion-based integrated circuit overlay error measurement method provided by the application specifically includes:

[0031] (1) According to the existing lithography mark manufacturing process in the semiconductor industry, the topographic features and optical constants of the material of the measured lithography mark are determined; for example, a typical lithography mark model is used for modeling here, and the specific topographic feature parameters are as follows: Si substrate, Si grating in the first layer, gap filled with SiO2 film, and PR grating in the second layer; the specific size parameters of each structure are as shown in Table 1. Figure 2

[0032] (2) The lithography mark model is simulated and analyzed by using the rigorous coupled wave analysis method (RCWA), and the topographic size parameters in step (1) are input into the RCWA algorithm, and the measurement configuration is determined, such as wavelength, azimuth angle, polarization of incident light, etc.

[0033] (3) After the spectrum information calculated by RCWA, the spectrum signal or Mueller matrix obtained by simulation is loaded with noise (ΔI or Δm) as the actual spectrum signal, and the noise is simplified to follow the Gaussian distribution N = (μ, σ 2 ), and then a large number of different measurement values OVL m are calculated by using a variety of extraction methods (MBO, EBO and their combinations) on the spectrum signal loaded with noise; according to the prior knowledge of the manufacturing process, the value limit of the overlay error is determined as ε max , if , the measurement value ε0 is not included in the fusion object.

[0034] (4) The actual measured spectrum signal is obtained, and a plurality of different measurement values OVL m are calculated by using the different extraction methods in step (3), and a new overlay error value OVL m is obtained by using information fusion technology; the information fusion technology has two stages: local fusion and overall fusion; the local fusion technology is reflected in that: in different overlay error extraction methods, for the MBO+MBO and MBO+EBO extraction method fusion measurement, respectively, a higher credibility overlay error value is obtained by using the information fusion method. For the results obtained by the MBO and EBO extraction methods alone, the average value is solved to obtain the overlay error value representing the extraction method

[0035] Next, the local fusion technology will be carefully introduced, taking the MBO+EBO fusion measurement extraction method as an example, and the specific steps are as follows:

[0036] ① Use the RCWA algorithm to simulate the mark, load the spectrum signal or Mueller matrix with Gaussian distribution N = (μ, σ 2 ) noise as the actual spectrum signal, and use the MBO and EBO methods to obtain different overlay error value sets OVL1 = {ε1, ε2, …, ε n} and​ And based on the value ranges of OVL1 and OVL2, the absolute value range of the overlay error is obtained as |ε|=[0,d];

[0037] ② Based on the fundamental concepts of DS evidence theory, the identification framework of the fusion process is a finite set. According to this principle, the absolute value range of the overlay error is discretized, and |ε| is divided into n equal parts and represented by the set Θ. This is the identification framework for all events. The method for determining the basic probability assignment of each event is as follows: first, determine the prior threshold δ (the distance to the focal element is within a certain range) based on the identification framework; then, based on the measured value ε... k With Jiao Yuan The distance between them is compared with δ, and the specific steps are as follows: When When k, j∈[0, n], it can be determined that This incident has happened once before; if If i events satisfy the above rules, then The basic probability distribution of the event is Based on this calculation rule, a basic probability assignment number is given to each possible overlay error value:

[0038]

[0039]

[0040] ③ From Dempster's combination rule in the DS evidence theory, we get:

[0041]

[0042]

[0043]

[0044] …

[0045]

[0046] ④ After obtaining the basic probabilities of the possible values ​​of each overlay error, the maximum confidence value principle model is adopted (i.e., the confidence value of the focal element is the largest and significantly greater than the confidence value of the second largest focal element). At the same time, thresholds μ1 and μ2 are preset to ensure that the confidence of the selected focal element is not too small and the uncertainty of the whole set is not too large, thus ensuring that the result has credibility and decision-making power. The specific steps are as follows:

[0047] Given the identification frame Θ, and the basic probability assignment function of m defined on Θ, then:

[0048]

[0049]

[0050] If yes:

[0051]

[0052] The decision result is epsilon k If the two maximum focal element confidence values are close, the decision proposition is further contracted or more accurate and richer evidence information is added to help the decision.

[0053] Meanwhile, the two results in the MBO+MBO extraction method are fused by using the above D-S evidence theory method to obtain a new overlay error value.

[0054] (4) In this embodiment, four different overlay error extraction methods of MBO, EBO, MBO+MBO and EBO+MBO are fused, four overlay error result sets OVL1, OVL2, OVL3 and OVL4 are obtained according to the above calculation method, and each extraction method is assigned a probability of 1 / 4, so the total is still the framework The probability distribution number of each event is 1 / 4, the four extraction method results are fused by using the above same evidence theory method, and a more accurate and reliable overlay error value OVL * is obtained.

[0055] In addition, the present application is not limited to the above-mentioned embodiments, and the above is only the preferred embodiment of the present application, and the protection scope of the present application is not limited to the above-mentioned embodiments, and any technical solution belonging to the idea of the present application belongs to the protection scope of the present application. For example, in the above-mentioned embodiments, the optical structure modeling solving algorithm is not limited to the RCWA algorithm, and methods such as finite element and time domain finite difference method can also be used; the information fusion technology is not limited to the D-S evidence theory, and methods such as Bayes inference or artificial neural network can also be used; if the D-S evidence theory is used, the final decision model is not limited to the maximum confidence value principle model, and the "minimum point" principle or the decision model based on the minimum risk can also be used.

Claims

1. A method for measuring overlay error of integrated circuits based on information fusion, characterized in that, Includes the following steps (1) Based on the relevant information of the process object to be tested, determine the morphological features and material properties of the overlay mark to be tested, and use the parametric modeling and electromagnetic field calculation method to calculate the optical response of a set of typical DBO overlay marks to obtain the optical signal under ideal conditions. (2) Add noise Δm to the simulated spectral signal ΔI or the Mueller matrix as the simulated measured spectral signal; (3) The above signals were extracted using model-based extraction methods, experience-based extraction methods, and a combination of both. Each extraction method calculated multiple different measured values ​​(OVL) for the noise-loaded spectral signal. m ; (4) Based on the results of batch simulation calculation, determine the reasonable range of values ​​for the overlay error distribution, and exclude actual measured values ​​that are not within the distribution range; (5) For different extraction methods in step (3), calculate the actual measurement signal to obtain multiple different measurement values ​​OVK. m A new overlay error value OVK is obtained by information fusion. m The information fusion process has two stages: local fusion and global fusion. Local fusion involves: for the fusion measurement results of MBO+MBO and MBO+EBO extraction methods, each method is used to obtain a more reliable overlay error value; for the results obtained from individual MBO and EBO extraction methods, the average value is used to obtain the overlay error value representing that extraction method. (6) As described in step (5), the overall fusion is reflected in: the OVL obtained by each extraction method m A secondary decision fusion is performed, that is, the results of several extraction methods such as MBO, EBO, MBO+MBO, and EBO+MBO are fused to obtain a more reliable overlay error value OVL. * .

Images